Everyone has an opinion about AI in production. Fewer people have a changelog. This session reports findings from a series of interviews with practitioners across virtual production, XR, CG animation, VFX and games. Supervisors, pipeline engineers, content leads, producers, conducted under one constraint: no futurism. Only what has actually been put into a pipeline, what it replaced, what it cost, and what was taken back out. Interviewing across five disciplines rather than inside one turned out to matter. These fields share a pipeline shape: generate, iterate, review, approve, deliver to a display, but hit its problems in a different order and at different times. Some of what feels new in one discipline is a solved problem two rooms over. The session covers where the tooling genuinely shortened a loop, where it moved cost downstream instead of removing it, where teams reverted, and which discipline got there first. It closes with where to put your next 100 hours, and where not to.
